We study the problem of large-scale network embedding, which aims to learn latent representations for network mining applications. Previous research shows that 1) popular network embedding benchmarks, such as DeepWalk, are in essence implicitly factorizing a matrix with a closed form, and 2) the explicit factorization of such matrix generates more powerful embeddings than existing methods. However, directly constructing and factorizing this matrix—which is dense—is prohibitively expensive in terms of both time and space, making it not scalable for large networks.
Qiu et al - NetSMF ~ Large-Scale Network Embedding as Sparse Matrix Factorization.pdf:C\:\\Users\\Admin\\Documents\\Research\\_Paperbase\\Graph Embeddings\\Qiu et al - NetSMF ~ Large-Scale Network Embedding as Sparse Matrix Factorization.pdf:application/pdf
%0 Conference Paper
%1 qiu_netsmf:_2019
%A Qiu, Jiezhong
%A Dong, Yuxiao
%A Ma, Hao
%A Li, Jian
%A Wang, Chi
%A Wang, Kuansan
%A Tang, Jie
%B The World Wide Web Conference on - WWW '19
%C San Francisco, CA, USA
%D 2019
%I ACM Press
%K Geometry_Study Matrix_Factorization Node_Embeddings Skip-Gram
%P 1509--1520
%R 10.1145/3308558.3313446
%T NetSMF: Large-Scale Network Embedding as Sparse Matrix Factorization
%U http://dl.acm.org/citation.cfm?doid=3308558.3313446
%X We study the problem of large-scale network embedding, which aims to learn latent representations for network mining applications. Previous research shows that 1) popular network embedding benchmarks, such as DeepWalk, are in essence implicitly factorizing a matrix with a closed form, and 2) the explicit factorization of such matrix generates more powerful embeddings than existing methods. However, directly constructing and factorizing this matrix—which is dense—is prohibitively expensive in terms of both time and space, making it not scalable for large networks.
%@ 978-1-4503-6674-8
@inproceedings{qiu_netsmf:_2019,
abstract = {We study the problem of large-scale network embedding, which aims to learn latent representations for network mining applications. Previous research shows that 1) popular network embedding benchmarks, such as DeepWalk, are in essence implicitly factorizing a matrix with a closed form, and 2) the explicit factorization of such matrix generates more powerful embeddings than existing methods. However, directly constructing and factorizing this matrix—which is dense—is prohibitively expensive in terms of both time and space, making it not scalable for large networks.},
added-at = {2020-02-21T16:09:44.000+0100},
address = {San Francisco, CA, USA},
author = {Qiu, Jiezhong and Dong, Yuxiao and Ma, Hao and Li, Jian and Wang, Chi and Wang, Kuansan and Tang, Jie},
biburl = {https://www.bibsonomy.org/bibtex/2686c6d732d435775ea7085430ddbcdc5/tschumacher},
booktitle = {The {World} {Wide} {Web} {Conference} on - {WWW} '19},
doi = {10.1145/3308558.3313446},
file = {Qiu et al - NetSMF ~ Large-Scale Network Embedding as Sparse Matrix Factorization.pdf:C\:\\Users\\Admin\\Documents\\Research\\_Paperbase\\Graph Embeddings\\Qiu et al - NetSMF ~ Large-Scale Network Embedding as Sparse Matrix Factorization.pdf:application/pdf},
interhash = {c69ce0aba77c8db37474d24636097c7d},
intrahash = {686c6d732d435775ea7085430ddbcdc5},
isbn = {978-1-4503-6674-8},
keywords = {Geometry_Study Matrix_Factorization Node_Embeddings Skip-Gram},
language = {en},
pages = {1509--1520},
publisher = {ACM Press},
shorttitle = {{NetSMF}},
timestamp = {2020-02-21T16:09:44.000+0100},
title = {{NetSMF}: {Large}-{Scale} {Network} {Embedding} as {Sparse} {Matrix} {Factorization}},
url = {http://dl.acm.org/citation.cfm?doid=3308558.3313446},
urldate = {2019-12-10},
year = 2019
}